Multi-level context-dependent acoustic modeling for automatic speech recognition | IEEE Conference Publication | IEEE Xplore